To improve the convergence speed of modeling for switched reluctance motor(SRM) using a-daptive network based fuzzy inference system (ANFIS), a mathematic model for SRM was proposed. This method combined the genetic algorithms (GA) with ANFIS. The fuzzy neural networks adopted Takagi-Sugeno model which has strong robustness and adaptability. CA were introduced to traditional hybrid algorithm of ANFIS to improve the training speed during parameter training. Based on testing sample data of 8/6 SRM, the inductance and torque models were established. The simulation results test that the new model has the merits of higher precision and strong generalization ability, and the convergence speed is improved more than twice. The SRM system were simulated with the trained inductance and torque models. Compared with the actual system, the current waves are similar. This proves the new modeling method is correct and feasible.%针对在使用自适应神经网络模糊推理系统对开关磁阻电机进行建模的过程中收敛速度慢的问题,采用将遗传算法和自适应神经网络模糊推理系统相结合的开关磁阻电机建模方法.网络结构仍然采用具有很强鲁棒性和自适应性的Takagi-Sugeno模型,而在网络参数训练时将遗传算法与自适应神经网络模糊推理系统的传统混合学习算法相结合,以提高训练速度.根据实测的8/6极开关磁阻电机的样本数据,对开关磁阻电机的电感和转矩进行建模,仿真结果表明,该方法具有很高的精确度和很强的泛化能力,并且将收敛速度提高了两倍多.将所建模型应用到开关磁阻电机控制系统仿真中,并与实际控制系统进行对比,两者结果基本一致,证明了该方法的正确性和可行性.
展开▼
